Depleted melt inclusions in MORB plagioclase: Messages from the mantle or mirages from the magma chamber?

Melt inclusions that are depleted in high field strength elements (HFSE; Ti, Zr, Nb), relative to other incompatible elements, were found in a plagioclase phyric normal mid-ocean ridge basalt (N-MORB) from the southern Mid-Atlantic Ridge. Similar inclusions are present in many other phyric NMORB. HFSE-depleted inclusions constitute only a few percent of all melt inclusions in this sample, and inclusions within individual crystals display a limited range of HFSE-depletion. Relative to host glass, they are depleted in the order: Nb<Zr<Ti ≈ HREE ≈ Th < LREE < U. Concentrations of Si, Al, Fe, Mg and Ca are similar to the host glass. Large ion lithophile elements (LILE) are enriched relative to the host glass in the order: Rb>Ba>K>Pb>Na>Sr. La/Sm is higher than in the host glass. C1 is enriched but not to the level observed in HFSE-depleted inclusions by Nielsen et al. [Geochem. Geophys. Geosyst. (2000) 1], who deemed similar inclusions in other MORB as "C1-enriched". HFSE depletion is not related to inclusion size, plagioclase host composition, or inclusions' Mg#s. Because of the disparate behavior of elements with similar bulk crystal-liquid partition coefficients, the depletion trends cannot be modeled by any process that involves crystal liquid equilibrium, such as melting or crystallization. Nielsen et al. proposed that similar inclusions represent liquids that were formed by melting of hydrothermally altered depleted peridotite. An alternative explanation is that the inclusions' compositions were controlled by diffusional processes. There is a good correlation of the elements' abundance relative to the host with Z²r_i, a quantity that is highly correlated with diffusion in silicate liquids ([Hofmann, A.W., 1980. Diffusion in natural silicate melts: a critical review. Physics of Magmatic Processes, Princeton Univ. Press, pp. 385-417]; Z = atomic radius and r_i = ionic radius) and possibly in plagioclase. The depletions are consistent with a model in which plagioclase rapidly dissolves to form a plagioclase-like melt, while diffusion through liquid channels or solid plagioclase transports elements from the host liquid to the inclusion. Plagioclase dissolution may have been aided by increased C1 + H₂O in the enclosing magma, which might have been ultimately derived from hydrothermal activity. A second diffusion-related model starts with the entrapment of ultradepleted melt inclusions such as those found in olivine. Subsequent diffusion from a less depleted host melt through solid plagioclase into the inclusion would control the inclusion's incompatible element abundances. Comparison of elements with similar Z²r_i (e.g., LREE vs. HREE) suggests that the host liquid of the analyzed sample was enriched in incompatible elements despite the inclusions' depletion in Nb, Zr and REE. Compositions of inclusions that are not HFSE-depleted also suggest that the host melt was enriched. Ultradepleted inclusions in MORB olivine [Nature 363 (1993) 151] are not like those in plagioclase: they are depleted in all incompatible elements, and their compositions are consistent with progressive fractional melts of the mantle (ibid). However, the possibility of a diffusion relationship should also be examined.